MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202611090964 A) filed by Noida Institute Of Engineering And Technology Niet on July 27, 2026, for System And Method For Adaptive Hypergraph Attention Based Electrocardiogram Classification With Uncertainty Quantification.
Inventors include Mr. Yaduvir Singh; and Dr. Vikas Sagar.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: A system (100) and method for classifying multi-lead electrocardiogram (ECG) signals using an adaptive hypergraph attention network with clinical context fusion and Bayesian uncertainty quantification is disclosed. The system acquires a multi-lead ECG signal, extracts P, Q, R, S and T landmark points with confidence scores, and constructs a hypergraph in which each hyperedge connects more than two landmark points across intra-lead, inter-lead and temporal dimensions. Weights of the hyperedges are learned adaptively rather than being pre-defined. Patient-specific metadata is injected as a virtual global node through a gated fusion unit. A Bayesian hypergraph attention network with variational dropout computes class probabilities and a Monte Carlo sampling engine estimates epistemic and aleatoric uncertainty, so that low-confidence predictions are rejected for clinician review. A quantised network is deployed on a neural processing unit for real-time inference on wearable devices, with an explainability module
Disclaimer: Curated by HT Syndication.
